Why AI Isn’t a Bubble – And Why That Changes Everything

The infrastructure is real, the money is real, and refusing to engage won’t protect you

Data deep dive — AI and data driven articles

Update: 11/20/2025 Employment Data Update
Latest employment data reveals a more complex picture than initial reporting suggested. While BLS reports 119,000 total jobs added and unemployment at 4.4%, continuing jobless claims rose to 1.974 million (up 28,000 which is slightly about consensus) meaning displaced workers are taking longer to find new employment.

The divergence: Healthcare, government, and hospitality continue hiring (low-automation roles), while tech, customer service, and administrative sectors shed jobs (high-automation roles). Goldman Sachs estimates 50,000 job losses in October specifically in automation-vulnerable sectors.

The critical signal: It’s not just that jobs are being lost, it’s that reemployment is getting harder. Initial claims average 227,000 per week, but continuing claims rising indicates workers struggle to find new positions after displacement. When automation eliminates a $75,000 data analyst role and the economy adds two $32,000 healthcare aide positions, statistics show “net job growth” while workers experience downward mobility.

Update: 11/19/2025 Based on Federal Reserve statements made today:
Federal Reserve’s Mixed Signals

Federal Reserve Chair Jerome Powell doesn’t think this is a bubble. In October 2025, he made it explicit: Unlike the dotcom era, today’s AI companies “actually have earnings. These companies actually have business models and profits.” When asked about AI’s economic impact: “The productivity gains from AI could be substantial and lasting… We’re seeing real capital investment, not speculation.”

However, recent Fed meeting minutes reveal internal concerns. Several Federal Reserve participants highlighted “the possibility of a disorderly fall in stock prices, especially in the event of an abrupt reassessment of AI-related prospects.”

What this tells us: The Fed acknowledges AI infrastructure spending is real and productive, but they’re watching stock valuations carefully. The physical infrastructure (data centers, chips, power grids) represents genuine economic value. The question is whether stock prices have run ahead of near-term fundamentals.

For workers, this distinction matters: Even if AI stock prices crash, the infrastructure being built is real and permanent. The data centers will still need operators. The automation will still happen. A stock market correction doesn’t stop the transformation. It just might slow the pace of new construction.

Remainder of the article is as was originally published 11/19/2025

Bottom Line Up Front

AI doesn’t care if you believe in it. It doesn’t need your approval to reshape your industry. And refusing to learn about it, understand it, or prepare for it doesn’t stop it. It just leaves you unprepared when it arrives.

Companies earning 55-75% profit margins on AI chips… Eliminating 50,000+ jobs per month… Paying executives record bonuses… Driving stock prices up 200-400%…

Are receiving BILLIONS in taxpayer subsidies to build the infrastructure that will automate away MORE jobs.

And there’s virtually no regulation preventing them from doing any of this.

This isn’t capitalism. This is wealth extraction with a government subsidy.

As of November 2025, we’re documenting over $1 trillion in confirmed AI infrastructure investments, with new multi-billion dollar commitments announced weekly. Meta: $600 billion through 2028. Microsoft: $80 billion in 2025 alone. Google: €5.5 billion in Germany, creating 9,000 jobs. Anthropic: $50 billion in new U.S. data centers with Microsoft/Nvidia backing. Amazon Web Services: $5.3 billion Saudi Arabia commitment. AMD/xAI: Multi-billion dollar Saudi Arabia partnership. AMD’s CEO says demand for AI compute infrastructure is “insatiable,” projecting $100 billion in annual chip revenue within five years while maintaining premium margins of 55-58%.

Federal Reserve Chair Jerome Powell doesn’t think this is a bubble. In October 2025, he made it explicit: Unlike the dotcom era, today’s AI companies “actually have earnings. These companies actually have business models and profits.” When asked about AI’s economic impact: “The productivity gains from AI could be substantial and lasting… We’re seeing real capital investment, not just speculation.” The Fed Chair doesn’t make bullish statements about technology trends. When he does, it’s because the capital flows are real enough to affect monetary policy.

Goldman Sachs agrees, calling the spending levels “sustainable.” The physical infrastructure being built – data centers, power grids, fiber networks – represents durable economic value, not speculation. In their September 2025 infrastructure report: “AI data center buildout represents the largest private infrastructure investment since the interstate highway system. $1+ trillion through 2030 in just data centers and power generation.”

Meanwhile, the U.S. lost 11,250 jobs in the latest weekly data, with Goldman Sachs estimating 50,000 jobs lost in October alone. Companies are explicitly using AI to reduce headcount with 11% actively cutting staff due to AI, rising to 31% in tech sectors.

Here’s the reality check nobody wants to hear but everyone needs to understand: You’re not choosing whether AI happens. That choice was made by trillion-dollar investments, government backing, and genuine productivity gains that companies can measure on their balance sheets. You’re choosing whether you see it coming or get blindsided by it.

The workers who saw automotive manufacturing leave Michigan decades ago and prepared themselves had options. The workers who insisted it would never happen lost everything. This is that moment again but bigger, faster, and backed by more money than automotive outsourcing ever was.

This isn’t about loving AI or thinking job displacement is good. It’s about survival. It’s about having options when change arrives. It’s about maintaining agency in your own life.

Why This Isn’t a Bubble: Following the Money Trail

Two Massive Infrastructure Projects (Not to Be Confused)

Before we dive in, let’s clarify the major projects:

Project Stargate = OpenAI/SoftBank/Oracle partnership, $500B through 2029
Meta’s AI Infrastructure = Meta’s company-wide buildout, $600B through 2028
(Note: Google’s “Project Suncatcher” is a completely separate moonshot exploring space-based data centers using satellites – not related to the ground-based buildout we’re discussing here but will be addressed in subsequent articles in this series.)

These are separate but parallel investments – they are happening simultaneously. Both are real, both involve physical data centers being built right now, and together they represent over $1 trillion in committed infrastructure spending.

The Reality Check: AI Doesn’t Need Your Approval

Here’s what makes this different from previous tech hype cycles: the money is already being spent, the physical infrastructure is already being built, and the government considers it a national security priority.

This isn’t 1999 when companies with no revenue were getting billion-dollar valuations. This is concrete and steel and chips being purchased right now. You and I may have a say in where, but the when and the if are not in question.

$1+ Trillion Being Built

Let’s put actual numbers to this buildout. These aren’t projections or aspirations. These are announced commitments with construction already underway or contracts already signed. Warning – lots of bullet point lists to follow. This analysis presents a lot of data and numbers driven information.

Historical Data Center Capital Expenditures:

2023 Baseline: ~$300 billion global data center spending
2024 Acceleration: $430-455 billion (51% increase in one year)
- Top hyperscalers (AWS, Microsoft, Google, Meta): ~$180 billion
- AI accelerated servers jumped from 15% to 35% of enterprise budgets
- Meta specifically: $38-40 billion
2025 Projection: $598 billion (30% additional growth)
- Meta alone: $60-65 billion (nearly doubling their 2024 spend)
- Microsoft: $80 billion
- Amazon: $100 billion
- Google: $75 billion
2029 Forecast: $1.1 trillion (2.5x the 2024 spend)

What this progression shows: This isn’t speculation or venture capital gambling. These are public companies with shareholders, earnings calls, and quarterly reporting requirements. They’ve already nearly doubled their infrastructure spending from 2023 to 2024, and they’re projecting to triple it by 2029.

The Chip Layer: Where Hype Meets Hardware

Want to know if AI infrastructure is real? Follow the semiconductor orders. You can’t fake chip manufacturing – it takes 3-6 months to produce advanced processors, requires signed contracts, and companies pre-pay billions. This infrastructure needs processors. The chip spending alone is staggering.

AMD’s Data Center Revenue Projections:

2024: $19 billion (actual)
2025: $30-35 billion projected (60-84% growth)
2026-2027: Analyst estimates suggest $50B+ annually
Projecting $100 billion in annual data center chip revenue within 5 years
$45 billion in custom chip design revenue starting 2026
Major contracts: OpenAI ($500M starting 2026), Oracle, Meta
55-58% gross margins on AI chips
Earnings per share projected to reach $20+ (300-400% growth)
Note: Revenue guidance excludes China due to export controls

Key AMD Validation Points:

Signed customer contracts: $500M OpenAI deal starting 2026, plus multi-billion dollar commitments from Oracle and Meta
“Very clear forecasts from cloud customers” (AMD CEO Lisa Su) = Not aspirational projections, backed by contracts
Wall Street consensus: 27 Buy ratings vs 10 Hold, with Bank of America projecting even higher than AMD’s own guidance
Premium margins: 55-58% gross margins on MI300 chips = real demand, not fire-sale pricing

NVIDIA’s Data Center Dominance:

Q4 2024: $22.1 billion data center revenue (90%+ GPU market share)
Premium margins: 74-75% gross margins
2025 guidance: $120B+ annual data center revenue
Supply constraints: Can’t make chips fast enough to meet demand
$115+ billion in data center revenue for fiscal 2025
$500+ billion in secured orders for Blackwell and Rubin GPUs
Fiscal 2026 data center revenue could reach $170 billion
74-75% gross margins
Jensen Huang: $3-4 trillion AI infrastructure market by 2030
Note: Guidance assumes no H20 chip shipments to China

Combined Picture:

AMD + NVIDIA combined: $270B+ in projected annual data center chip revenue
Both companies showing 55-75% profit margins = pricing power from real demand
Combined backlog: $500B+ in secured orders through 2027
Both explicitly excluding China market due to export controls (see below)

Export Controls Prove This Is Strategic

U.S. government semiconductor export restrictions to China (implemented November 2024) continue driving U.S. companies toward alternative markets. Recent November 2025 partnerships with Saudi Arabia demonstrate this strategic shift.

Both AMD and NVIDIA 2025 guidance explicitly excludes China market:

AMD: “Our guidance assumes minimal China revenue”
NVIDIA: “Assumes no shipments of H20 chips to China”

Translation: These companies are projecting massive growth without their largest potential customer. The U.S. government considers AI chip manufacturing capacity a national security priority – that’s not bubble behavior.

Physical Infrastructure Being Built Right Now – The Infrastructure Layer

Meta Platforms:

$600 billion total AI infrastructure investment through 2028 (company-wide)
$60-65 billion in capital expenditures for 2025 alone
El Paso, Texas data center: $1.5 billion (initial phase), 1,800 construction jobs, 100 operational jobs
This is Meta’s 29th data center globally (3rd in Texas)
Part of broader AI infrastructure expansion, not a single named project

Microsoft:

$80 billion for AI-enabled data centers in fiscal 2025 (through June 2025)
Over half of spending in the U.S.
$10+ billion for AI infrastructure expansion in Portugal
“Super factory” sites for accelerated construction
Atlanta data center connected to Wisconsin facility forming a “massive supercomputer”

Google (Alphabet):

€5.5 billion ($5.9 billion) for Germany cloud infrastructure
New data center in Dietzenbach, Germany creating 9,000 jobs
Part of broader global data center expansion
Project Suncatcher: Separate moonshot exploring space-based AI data centers powered by solar satellites (prototype launch 2027) – unrelated to ground-based buildout

OpenAI – Project Stargate:

$500 billion partnership with SoftBank and Oracle through 2029
First $100 billion committed for 2025-2026
Five data center locations:
- Project Jupiter: Santa Teresa, New Mexico – $165 billion
- Abilene, Texas (construction started)
- Austin area, Texas
- Ohio
- Undisclosed Midwest location
Building physical AI infrastructure for OpenAI’s models

Oracle:

Partner in Project Stargate
Multiple multi-billion dollar cloud deals
$60+ billion in “Neocloud” commitments

You can’t fake construction permits, power grid upgrades, and land purchases. This is physical infrastructure with real capital commitments and construction employment already happening.

The Difference From Crypto/NFT Bubbles

Crypto/NFT Bubble Characteristics:

No underlying productive use
Purely speculative assets
No physical infrastructure
Minimal employment creation
Collapsed when sentiment shifted

AI Infrastructure Reality:

Massive productivity gains in code generation, customer service, content creation (measurable business value)
Physical data centers, chip fabrication plants, power generation facilities
Hundreds of thousands of construction and operational jobs
Federal Reserve and Goldman Sachs validation
Government considers it strategic national priority

The Power and Grid Infrastructure (The Hidden Cost)

Data centers need massive amounts of electricity. This requires parallel investment in power infrastructure:

New substations and transmission lines
Grid capacity upgrades
In some cases, dedicated nuclear or natural gas power plants
DTE Energy (Michigan): Negotiating deals for up to 7 gigawatts of data center power
- That’s more than half of DTE’s entire current grid capacity (11 GW)
- DTE Energy:
  - 7 GW data center pipeline
  - Current capacity: 11 GW total
  - Already delivering 9.5 GW peak
  - The 7 GW would exceed total capacity
  - Saline Township (OpenAI/Oracle/Related): 1.4 GW alone = 25% of DTE’s entire current load
  - Uses as much power as “750,000 homes” or “equivalent to the entire city of Detroit”
- Consumers Energy:
  - Current capacity: 7.6 GW
  - Data center pipeline: 15 GW in discussions
  - Would double their capacity

The Total Picture

Conservative estimate for 2025-2028:

Infrastructure (data centers, facilities): $1+ trillion
Chips and processors: $300+ billion
Power and grid upgrades: $100+ billion
Total visible commitments: $1.4+ trillion

And this doesn’t include:

Networking equipment (fiber, switches, routers)
Cooling systems and water infrastructure
Storage systems
Thousands of smaller projects not making headlines
The entire supply chain supporting this buildout
Conclusion: the impact is massive already

What This Means in Context

To put $1.4 trillion in perspective:

Larger than most countries’ entire GDP
The combined GDP of Japan and Germany: ~$7 trillion
McKinsey estimates companies will invest almost $7 trillion in global data center infrastructure by 2030
U.S. GDP: ~$29 trillion
This represents roughly 5% of U.S. GDP being invested in AI infrastructure over 4 years

JPMorgan economists project that AI-related infrastructure spending could add 0.2 percentage points to U.S. GDP growth over the next year, roughly the same boost that shale drilling delivered at its peak.

Bottom Line

This isn’t a bubble. This is a capital deployment cycle similar to:

Railroad expansion (1860s-1890s)
Electrification (1900s-1930s)
Interstate highway system (1950s-1980s)
Internet backbone infrastructure (1990s-2000s)

The Jobs Math

These aren’t just financial numbers. They represent real construction and operational jobs:

Meta El Paso: 1,800 construction, 100 operational
Google Germany: 9,000 jobs secured
Project Jupiter (Stargate): 750 permanent jobs, thousands of construction jobs
Multiply across hundreds of projects nationwide

But here’s the tension: While data center construction creates thousands of jobs, AI deployment is eliminating tens of thousands in other sectors. More on that shortly.

The Acceleration

Perhaps most importantly: This spending is accelerating, not slowing.

2023: ~$300 billion baseline spending
2024: $430-455 billion (51% increase)
2025: Projected $598 billion (30% additional growth)
2026-2028: Sustained or growing toward $1.1 trillion by 2029

AMD CEO Lisa Su: “Data center growth to be faster in the near term.”

This is the opposite of a bubble deflating. This is exponential growth in committed capital.

The wealth being created is real. The question is: who captures it?

Spoiler: It’s not the workers building the data centers.

Who Captures the Wealth: The Margins vs. Wages Disconnect

The infrastructure spending is real. The chip orders are real. The profits are real. But here’s the uncomfortable question nobody in Silicon Valley wants to answer: Where is all that money going?

Let’s look at one company to understand the wealth distribution: AMD.

The AMD Case Study: Following the Money

AMD’s AI Data Center Projections:

Current annual revenue: ~$25-30 billion (2024)
Projected data center revenue: $100 billion annually within 5 years
Gross margins on AI chips: 55-58% (premium pricing, no competition pressure)
Current earnings per share: ~$5-6
Projected EPS in 3-5 years: $20+
That’s a 300-400% increase in earnings per share

What This Means in Dollar Terms:

At 55-58% gross margins on $100 billion in annual revenue, AMD will generate $55-58 billion in gross profit per year from AI chips alone.

Who Benefits From This:

✅ AMD Shareholders: 300-400% earnings growth over 3-5 years
✅ AMD Executives: Compensation packages tied to stock performance
✅ Wall Street: Trading fees, investment banking fees, fund management fees
✅ Venture Capitalists: Early investors in AI companies see massive returns

Who Doesn’t Benefit:

❌ Workers displaced by AI: Zero compensation for job loss
❌ Communities hosting data centers: Tax breaks mean minimal revenue despite resource sacrifice
❌ Workers whose productivity AI enhances: Gains captured by employers, not workers
❌ Entry-level workers: Positions eliminated before they can enter workforce

It’s Not Just AMD

NVIDIA’s Story:

Gross margins: 74-75% on AI chips
Data center revenue: $115+ billion (fiscal 2025)
That’s $85+ billion in gross profit annually
Stock up over 200% in 2 years
CEO Jensen Huang’s net worth: $100+ billion

Microsoft, Google, Meta:

All reporting record profits
All deploying AI to reduce headcount
All seeing stock prices surge
All paying executives record bonuses

Meanwhile, American Workers

Are losing jobs at an accelerated rate.

October 2025 Job Losses (Goldman Sachs Data):

50,000+ jobs lost in a single month
11,250 workers losing jobs in a single week
Concentrated in:
- Customer service (AI chatbots replacing humans)
- Data entry (automation)
- Content moderation (AI filtering)
- Basic coding (AI code generation)
- Entry-level analysis (AI data analysis)

The Math:

AMD shareholders: +300-400% returns projected
NVIDIA shareholders: +200% returns (already realized)
U.S. workers: -50,000 jobs/month projected

The Productivity-Wages Disconnect

Here’s how this plays out in practice:

Traditional Model:

Company invests in new technology
Productivity increases 20-40%
Company revenue grows
Profits shared: some to shareholders, some to workers through raises
Everyone benefits from growth

AI Model (Current Reality):

Company deploys AI → Productivity increases 20-40%
Company revenue grows → Margins expand (like AMD’s 55-58%)
Shareholders benefit → Stock prices rise
Executives benefit → Bonuses tied to efficiency gains
Workers get: Layoffs, wage stagnation, or “you should be grateful to still have a job”

The Numbers Don’t Lie

What Companies Are Reporting:

AMD: 55-58% margins (premium pricing power)
NVIDIA: 74-75% margins
Tech sector overall: 15%+ revenue growth (highest of any sector)
S&P 500 companies: Record profit margins

What Workers Are Experiencing:

Real wages: Stagnant or declining when adjusted for inflation
Job security: Declining as AI deployment accelerates
Benefits: Being cut as companies optimize costs
Entry-level opportunities: Disappearing as AI handles junior tasks

The Historical Parallel

This has happened before.

First Industrial Revolution (1760-1840):

Massive productivity gains from mechanization
Factory owners accumulated enormous wealth
Workers got: 16-hour days, dangerous conditions, child labor, poverty wages
It took decades of labor organizing, strikes, and legislation to force wealth sharing
Result: 8-hour workday, minimum wage, workplace safety laws, child labor protections

We’re at that inflection point again.

The technology creates real value. AMD’s 55-58% margins prove it. The question is whether workers will organize to demand their share, or whether the gains flow entirely to capital.

The Uncomfortable Truth

AI isn’t reducing the value of labor because workers are less productive.

AI is reducing the value of labor because capital now has an alternative. When companies can automate your job, they have leverage. They can offer lower wages, worse conditions, fewer benefits – because the alternative isn’t another worker, it’s a system that works 24/7 and never asks for a raise.

This fundamentally changes the employer-worker relationship.

And right now, the wealth created by that change is flowing almost entirely to shareholders.

What the Data Shows

Let’s be specific about where AI’s wealth is going:

Wealth Captured by Capital:

AMD: $55-58B annual gross profit (55-58% margins)
NVIDIA: $85B+ annual gross profit (74-75% margins)
Microsoft, Google, Meta: Combined hundreds of billions in market cap gains
Tech sector investors: Massive returns from AI deployment

Wealth Captured by Labor:

Workers displaced by AI: $0 (job loss)
Workers whose roles AI enhances: Minimal wage gains despite 20-40% productivity increases
Entry-level workers: Blocked from workforce as AI eliminates traditional entry points

The gap is widening, not narrowing.

Why This Matters for Your Career

Understanding this wealth distribution is critical because it reveals three truths:

Truth #1: The money exists. AMD’s 55-58% margins prove companies are making massive profits from AI. This isn’t about scarcity, there’s plenty of wealth being created.

Truth #2: You’re not imagining the disconnect. If you feel like productivity gains aren’t translating to better wages or conditions, you’re right. The data confirms it. The gains are being captured by shareholders, not workers.

Truth #3: This is a political choice, not economic inevitability. The wealth COULD be shared more broadly. Companies COULD use productivity gains to raise wages, shorten hours, or improve conditions. They’re choosing not to. Because right now, workers have no leverage to demand it.

The Two-Track Strategy You Need

Track 1 – Personal Positioning: Position yourself in roles where you capture some of AI’s gains:

Data center jobs (temporary but well-paid)
AI implementation roles (helping businesses deploy AI)
Coordination work AI can’t automate
Roles managing AI systems

Track 2 – Collective Action: Support policies that force wealth sharing:

Tax AI profits to fund transition programs
Require profit-sharing when companies deploy AI to eliminate jobs
Mandate retraining funded by companies deploying automation
Shorter work weeks as productivity rises
Universal basic income funded by automation gains

You need both tracks. Personal positioning protects your individual career. Collective action ensures the broader system is fair.

And They’re Getting Taxpayer Subsidies to Do It

Here’s the part that should make your blood boil: These enormously profitable companies are getting billions in taxpayer subsidies to build the infrastructure that will eliminate your job.

Meta’s El Paso Deal:

80% property tax abatement for 35 years
$12.5 million from the city for road infrastructure improvements
If Meta builds all five planned phases: $550 million in total tax breaks
Total jobs created: 100 permanent positions
Cost per job: $5.5 million in tax breaks

The National Picture:

States Losing $100+ Million Annually (2024-2025):

Texas: $1 billion in 2025 (revised up from $130 million projection just 23 months earlier)
Virginia: $732-928 million annually
Illinois: $370 million
Georgia: $296 million
Iowa, Nevada, Ohio, Minnesota, Washington, Tennessee: $100+ million each

Total disclosed losses: At least $6 billion over the past five years in just 16 states that report data.

The real number? Far higher. 12 out of 32 states with data center subsidies don’t even disclose how much they’re giving away.

The Job Creation Myth

What they promise: “Economic development,” “job creation,” “tax base growth”

What actually happens:

Construction jobs: 1,688 workers average (temporary, 2-3 years)
Permanent jobs: 30-157 workers average per facility
Return on investment: States get back 48-52 cents for every dollar in tax breaks they give

Example: One Microsoft data center in Illinois:

Received: $38 million in tax exemptions
Permanent jobs created: 20
Cost per job: $1.9 million

Compare This to the Robber Baron Era

I felt this was worse than the robber baron days. Data says, I was right.

Gilded Age (1870s-1900s):

Railroads got massive land grants and subsidies
BUT: They built infrastructure that employed millions
BUT: They created permanent jobs (track maintenance, stations, operations)
BUT: They faced regulation (Interstate Commerce Act 1887, Sherman Antitrust Act 1890)

AI Era (2020s-present):

Tech companies get billions in tax breaks
They build infrastructure that eliminates jobs
They create 30-50 permanent positions per $1+ billion investment
Zero meaningful federal regulation of AI deployment or labor displacement
NDAs hide the details from public scrutiny

The Secrecy

It’s even worse than it appears because companies hide behind NDAs and shell corporations:

Example – Tucson, Arizona:

Proposed data center known only as “Project Blue”
Residents and elected officials didn’t know it was Amazon until a document was accidentally sent to a journalist
After the vote had already happened

Example – Indiana:

Company called “Hatchworks” applied for tax exemptions
After receiving the subsidy, a filing revealed it was a Google subsidiary

Why the secrecy? Because if residents knew who was getting their tax dollars and how few jobs would be created, they’d oppose it.

The Utility Rate Scam

But wait, there’s more! Regular customers are subsidizing data centers’ electricity:

El Paso Example:

Meta’s facility will require massive power infrastructure upgrades
New substations, transmission lines, grid capacity
Who’s paying? El Paso Electric guaranteed in writing to prove Meta isn’t being subsidized by other customers
Translation: They had to explicitly promise NOT to do what utilities normally do (make everyone pay for infrastructure that benefits one customer)

Harvard Law School Research (March 2025): “Some U.S. utilities are forcing the public to pay for infrastructure designed to supply a handful of exceedingly wealthy corporations.“

Michigan Example:

DTE Energy:
- 7 GW data center pipeline
- Current capacity: 11 GW total
- Already delivering 9.5 GW peak
- The 7 GW would exceed total capacity
- Saline Township (OpenAI/Oracle/Related): 1.4 GW alone = 25% of DTE’s entire current load
- Uses as much power as “750,000 homes” or “equivalent to the entire city of Detroit”
Consumers Energy:
- Current capacity: 7.6 GW
- Data center pipeline: 15 GW in discussions
- Would double their capacity
- That’s more than half their entire current grid capacity (11 GW)
- Required infrastructure: Hundreds of millions in upgrades
Who pays? Likely spread across all ratepayers
The Ratepayer Issue:
- Michigan passed law prohibiting residential customers from shouldering data center costs
- BUT: Advocates fear loopholes – gas plants have 30-year lifespan, protection only lasts 15 years
- Inside Climate News: “Residents may end up paying” after protection period expires
- DTE trying to fast-track approval, skip public hearings
- Michigan AG Dana Nessel intervened: “Public hearing is the only way to ensure transparency”

The Industry Is Virtually Unregulated

Unlike the Robber Baron Era, there is very limited regulatory framework for AI deployment with a few state exceptions:

What existed in the Gilded Age:

Interstate Commerce Commission (1887) – Regulated railroads
Sherman Antitrust Act (1890) – Prevented monopolies
State railroad commissions – Monitored rates and practices
Labor organizing (eventually led to regulations)

What exists now for AI:

Limited regulation of AI deployment and worker displacement:
Federal Level:
- No comprehensive AI legislation
- No agency overseeing AI-related job displacement
- No mandatory retraining programs
- No requirements to share productivity gains with workers
State Efforts (Emerging):
State regulations focus on HOW AI makes decisions (preventing bias), not on WHETHER companies can eliminate jobs wholesale or what they owe displaced workers.
- New York:
  - NYC Local Law 144 (effective July 2023): Requires annual third-party audits of automated employment decision tools for bias; employers must notify candidates
  - NY AI Act (proposed, effective 2027): Addresses algorithmic discrimination in employment, includes private right of action
  - NY AI Consumer Protection Act (proposed, effective 2027): Risk management requirements for high-risk AI systems
  - LOADinG Act (2024): First-in-nation law limiting state agency use of AI in decision-making
  - NY Workforce Stabilization Act (proposed): Would require impact assessments before AI deployment; 2% surcharge on companies with 15+ employees displaced by AI
- California:
  - AB 2013: Requires AI impact disclosures for large companies
  - Right of publicity protections for AI-generated content
- Colorado:
  - AI Act: Designates employment as “high-risk” sector; regulates high-risk AI systems
- Illinois:
  - Biometric privacy protections (BIPA)
- Connecticut:
  - S.B. 2 (proposed): Impact assessments, algorithmic discrimination protections
- Texas:
  - Texas Responsible AI Governance Act (TRAIGA, proposed): Risk management policies, semi-annual impact assessments
- New Jersey:
  - Multiple proposed bills regulating AI in hiring
However: These state regulations focus primarily on bias and discrimination in AI systems, NOT on job displacement or wealth distribution. They address HOW AI makes decisions, not whether companies can use AI to eliminate jobs entirely.
No worker protection requirements when AI eliminates jobs
No mandatory retraining programs funded by companies
No limits on how fast companies can automate
No requirements to share productivity gains with workers
States competing to give away tax dollars

The result: Companies like AMD and NVIDIA capture 55-75% profit margins, eliminate tens of thousands of jobs, pay executives massive bonuses, and receive billions in taxpayer subsidies to do it.

Good Jobs First Assessment

The nonprofit research group that tracks corporate subsidies:

“We know of no other form of state spending that is so out of control. We recommend that states cancel their data center tax exemptions. Such subsidies are absolutely unnecessary for an extremely profitable industry dominated by some of the most valuable corporations on earth, such as Amazon, Microsoft, Apple, Meta, and Alphabet.”

Bottom Line on Subsidies

Let’s be crystal clear about what’s happening:

Companies earning 55-75% profit margins on AI chips… Eliminating 50,000+ jobs per month… Paying executives record bonuses… Driving stock prices up 200-400%…

Are receiving BILLIONS in taxpayer subsidies to build the infrastructure that will automate away MORE jobs.

And there’s virtually no regulation preventing them from doing any of this.

This isn’t capitalism. This is wealth extraction with a government subsidy.

Data Center Opportunity: Real Jobs, Real Money, Honest Trade-offs

After everything we’ve covered—the wealth flowing to shareholders, the taxpayer subsidies, the job losses—here’s the uncomfortable reality: There are real jobs being created in data center construction and operations. And they pay well.

This is the tension at the heart of the AI buildout: short-term opportunity exists alongside long-term displacement.

Let’s be honest about both.

The Real Numbers

Construction Phase (2025-2028 peak):

Wages: $25-45/hour depending on trade and location
Annual income: $52,000-$93,600 (full-time)
Duration: 2-3 years per facility
Total positions: Tens of thousands across multiple projects

Data Center Operator/Technician (Operational Phase):

Entry-level: $55,000-$75,000 annually
Experienced: $75,000-$95,000 annually
Senior/Specialized: $95,000-$130,000 annually
Benefits: Typically excellent (healthcare, 401k, paid time off)

Specific Examples:

Meta El Paso: 1,800 construction jobs at peak, 100 operational jobs
Project Jupiter (Stargate): 750 permanent jobs, thousands of construction jobs
Google Germany: 9,000 jobs secured
Microsoft nationwide: Thousands of positions across multiple facilities

The Automation Timeline

Here’s what the industry itself says about the future:

Uptime Institute Survey (2024):

47% of data center operators plan to reduce hands-on staff within 5 years
Automation focus: Routine maintenance, monitoring, basic troubleshooting
Roles most at risk: Junior technicians, entry-level operators

Industry Projections:

2025-2027: Peak hiring (infrastructure buildout)
2028-2030: Automation deployment accelerates
2030-2032: Significant reduction in hands-on operational staff

Translation: You have a 3-7 year window for most operational roles before automation reduces headcount.

What This Means For You

If you’re considering data center work, go in with eyes open:

✅ Good reasons to pursue it:

Immediate income: Jobs available now, good pay, solid benefits
Bridge opportunity: Use the 3-7 years to save money, gain skills, plan next move
Technical skills: Experience with industrial systems, HVAC, electrical, networking
Security clearances: Some facilities require clearances that have value elsewhere
Foot in the door: Path to AI infrastructure operations (managing the automation)

❌ Bad reasons to pursue it:

20-year career plan: The role won’t exist in its current form
Avoiding reskilling: This IS the reskilling—you’ll need another one after
Ignoring automation: Pretending it won’t happen doesn’t stop it
Only option thinking: If this is literally your only path, take it—but plan the next one

The Strategic Approach

If you take a data center job, treat it as a strategic 3-7 year play:

Years 1-2: Establish and Save

Max out 401k contributions
Build 6-12 month emergency fund
Pay down high-interest debt
Live below your means (resist lifestyle inflation)
Learn everything about the systems

Years 3-4: Position and Pivot

Identify which roles are being automated vs. which are managing automation
Get certifications in AI infrastructure management if that interests you
Otherwise, start researching your next career move
Network with people in sustainable roles
Continue saving aggressively

Years 5-7: Transition

Begin your planned transition before you’re forced out
Use savings as runway for career change
Either move into AI infrastructure operations (if available) or execute exit plan
Don’t wait until automation announcements start

What “Managing the Automation” Looks Like

Some data center roles will persist—they’ll just manage AI systems instead of doing hands-on work:

AI Infrastructure Operations Roles:

Monitoring AI-driven maintenance systems
Overseeing automated troubleshooting
Managing predictive failure systems
Coordinating between AI systems and physical infrastructure
Exception handling (things AI can’t solve)

Key difference: These roles require understanding both the physical systems AND the AI managing them. Start learning now if this interests you.

The Geographic Reality

Not all data centers are created equal:

Texas (El Paso, Abilene, Austin area):

Multiple projects simultaneously
Lower cost of living
No state income tax
Hot job market right now

Virginia (Northern Virginia):

Largest data center hub globally
Higher pay but much higher cost of living
Mature market (less construction, more operations)
Heavy automation focus

Midwest (Iowa, Ohio, Indiana):

Growing rapidly
Moderate cost of living
Manufacturing workers with transferable skills in demand
Watch for oversupply

The Cost-Benefit Analysis

What you gain:

$75,000-$95,000 annually for 3-7 years = $225,000-$665,000 total
Technical skills that transfer to other industries
Benefits and stability during employment
Time to plan next move while earning

What you risk:

Job obsolescence within 7 years
Industry-specific skills that may not transfer well
Geographic relocation that may be hard to reverse
Potential wage stagnation as automation looms

The math: If you save aggressively, you could bank $100,000-$200,000 over 5 years. That’s a significant runway for a career transition.

Who Should Seriously Consider This

Strong candidates:

Displaced manufacturing workers:

Transferable skills: Mechanical, electrical, HVAC, industrial systems
Familiar with shift work and physical environments
Often have relevant certifications already
Example: Automotive technician → Data center technician (similar systems, better pay)

Military veterans:

Security clearances (valuable for government data centers)
Technical training in electronics, systems, infrastructure
Comfortable with structured environments
Strong path from military technical roles to data center operations

Career transitioners with 5-10 year horizon:

Planning a major career shift but need income now
Can use data center income to fund retraining
Example: Teacher planning to transition to instructional design—work data center while getting certifications

Strategic savers:

Willing to live frugally and bank most of income
Have a specific financial goal (pay off debt, save for business, fund education)
View this as temporary but lucrative

Who Should Think Twice

Poor fit candidates:

Young workers (18-25) looking for career foundation:

You have 40+ years of work ahead
Starting in a role with 7-year horizon is risky
Better to invest those years in future-proof skills
Exception: If you’re using it to fund education in another field

Workers 10+ years from retirement:

The timeline doesn’t work in your favor
You need stable employment through retirement
Automation will hit before you can retire
Exception: If you have strong financial cushion and can retire early if needed

Anyone who will depend on this job long-term:

If you can’t afford to transition in 5-7 years, this is risky
If you’re not a strategic saver, you’ll end up in same position when automation hits
If you have no backup plan, you’re just delaying the inevitable displacement

The Honest Assessment

This is NOT:

A 20-year career path
A solution to AI displacement
Evidence that “AI creates as many jobs as it eliminates”

This IS:

A temporary opportunity created by infrastructure buildout
A potential 3-7 year income source with decent pay
A strategic option for specific situations
Real work with real paychecks—while it lasts

The Question You Need to Answer

“Can I use this opportunity to position myself for what comes next, or will it just delay the inevitable?”

If your answer is the former: Pursue it strategically with a clear exit plan.

If your answer is the latter: Look for longer-term opportunities instead.

Current Job Availability

Where to look right now:

Indeed/LinkedIn searches:

“Data center technician” + [your city/state]
“Data center construction” + [region]
“Critical facilities technician”
“Data center operations”

Companies actively hiring:

Meta (Texas facilities)
Microsoft (multiple states)
Google (various locations)
Amazon Web Services (nationwide)
QTS, CoreSite, Equinix (major data center operators)
General contractors: Hensel Phelps, JE Dunn (construction phase)

What they’re looking for:

HVAC certification
Electrical experience
Mechanical systems knowledge
Network infrastructure experience
Willingness to work shifts (24/7 operations)
Some require security clearances

The Skills That Transfer

If you work data center operations, these skills have value elsewhere:

✅ Will transfer:

Industrial HVAC systems (any facility with climate control)
Electrical systems (commercial buildings, industrial facilities)
Emergency power systems (hospitals, critical infrastructure)
Building management systems (any modern facility)
Safety protocols and compliance
Preventive maintenance planning

❌ Won’t transfer easily:

Data center-specific monitoring tools
Proprietary systems unique to hyperscale facilities
Some networking skills (unless you go deep into networking as a career)

The Financial Strategy

If you take a data center job, here’s your financial playbook:

Mandatory savings targets:

Year 1: Save 20-30% of gross income
Years 2-3: Increase to 30-40% if possible
Years 4-5: Maintain 30-40% savings rate
Years 6-7: If still employed, save 40-50% (you’re on borrowed time)

What this looks like:

$75K salary, 35% savings rate = $26,250/year saved
Over 5 years: $131,250 (before investment growth)
That’s 18 months of living expenses at $75K income level
Or 2+ years at a lower salary during career transition

Goal: Exit with enough savings to fund 12-18 months of a career transition or retraining program.

Combining with Other Strategies

Data center work pairs well with:

Evening/weekend upskilling:

Online courses while employed
Certifications in your next field
Building a side business slowly
Developing portfolio in another area

Geographic arbitrage:

Work in Texas (no state income tax), save aggressively
Relocate to home state for next career phase with savings intact

Debt elimination:

Use higher income to clear student loans, credit cards
Become debt-free before next transition
Reduces financial pressure during career change

The Exit Strategy Examples

Example 1: Manufacturing to Data Center to Skilled Trade Business

Year 0: Laid off from automotive plant
Years 1-5: Data center technician, save $120K
Year 5: Launch HVAC business using data center experience + savings
Outcome: Self-employed, skills-based business, no boss

Example 2: Data Center to AI Infrastructure Operations

Years 1-2: Entry-level data center operator
Years 2-4: Learn AI monitoring systems, get certifications
Years 4-5: Transition to AI infrastructure ops role (managing automation)
Outcome: Stayed in industry but shifted to more sustainable role

Example 3: Data Center to Complete Career Change

Year 0: Data center technician
Years 1-4: Save $100K, take evening coding bootcamp
Year 4: Junior developer role (lower pay but sustainable)
Years 5-10: Build software development career
Outcome: Used data center as bridge to different field

The Community Implications

If data centers are coming to your community:

Short-term benefits:

Construction jobs (real, but temporary)
Some permanent positions
Housing demand spike
Local business boost during construction

Long-term costs:

Massive tax breaks (lost revenue for decades)
Strain on power and water infrastructure
Very few permanent jobs relative to tax breaks given
Residents may subsidize utility costs

Net result: Usually negative for the community unless they negotiate better terms (see Section VII).

What This Means for our “Under the Radar” Strategy

Data center jobs ARE an “Under the Radar” opportunity, but with a critical caveat:

✅ We will include them because:

Real jobs available now
Good pay for 3-7 years
Can be strategic financial move
Not everyone knows about the opportunity window

⚠️ BUT We are warning readers because:

Not a long-term solution
Requires exit strategy
Automation timeline is real
Need to save aggressively

The positioning: “This is a sprint, not a marathon. Plan accordingly.”

Bottom Line

Data center jobs are real. The pay is real. The opportunity window is real.

So is the automation timeline.

If you pursue this path:

Go in with a 5-year mindset, not a 20-year career plan
Save 30-40% of your income
Start planning your exit from day one
Either transition into AI infrastructure operations or have a clear next move
Don’t let good pay now make you complacent about automation later

This can be a smart strategic move. If you treat it strategically.

It’s a terrible life plan if you pretend the automation isn’t coming.

Breaking: Even as This Publishes

The evidence keeps mounting.

Saudi Arabia’s AI Infrastructure Deals (November 2025)

While American workers watch jobs disappear, US tech companies are committing billions to build AI infrastructure overseas:

Amazon Web Services:

$5.3 billion commitment to Saudi Arabia data center infrastructure
Part of Saudi Arabia’s broader AI transformation partnership

AMD/xAI Partnership:

Major semiconductor and AI infrastructure deal
xAI (Elon Musk’s AI company) partnering with Saudi Arabia
US chip manufacturing supporting overseas AI buildout

The Pattern: US companies are choosing global expansion over domestic job creation. The infrastructure spending is real—it’s just not creating American jobs.

Microsoft-Nvidia-Anthropic Partnership (November 2025)

Anthropic (maker of Claude AI) locked in massive Azure compute capacity through multi-billion dollar Microsoft partnership:

What This Means:

Anthropic’s $50 billion data center plan now has Microsoft/Nvidia backing
Confirmed compute capacity = sustained infrastructure demand
Competition between OpenAI, Anthropic, Google driving parallel buildouts

The Validation: When AI startups sign multi-billion dollar compute contracts with 3-5 year commitments, they’re betting their company’s survival on sustained infrastructure needs. This isn’t speculation—it’s existential for them.

Enterprise Adoption Accelerating

Target Corporation (400,000+ employees) publicly confirmed ongoing use of ChatGPT Enterprise for operations automation.

The Hidden Scale:

70% of Fortune 500 companies now use Microsoft Copilot
Thousands of enterprises have ChatGPT Enterprise
Every major retailer, bank, and consulting firm deploying AI

What workers don’t see: Your company is probably already using enterprise AI. The automation is happening now, quietly, department by department. By the time you notice, it’s already deployed.

Google Gemini 3 Crushes Benchmarks

November 2025: Google’s Gemini 3 AI outperformed OpenAI and Anthropic in business operations benchmarks.

Why This Matters:

Competition is accelerating, not slowing
Each company needs MORE infrastructure to stay competitive
Business automation use cases are proven and expanding
The race is real—which means sustained demand is real

The Timeline

These announcements came while this article was being finalized:

Saudi deals: Week of November 11-17
Microsoft-Anthropic: November 14
Target confirmation: November 18
Gemini 3 benchmarks: November 18

The buildout isn’t slowing. It’s accelerating.

Even as workers lose 50,000 jobs per month, companies are announcing new multi-billion dollar partnerships, locking in compute capacity for years, and expanding globally.

This is the opposite of a bubble deflating.

All sources for breaking developments documented in our comprehensive sources list.

Conclusion: The Reality You Can’t Ignore

The AI infrastructure buildout is real. The spending is real. The profits are real.

This isn’t a bubble waiting to burst. This is a $1+ trillion capital deployment already underway, with physical data centers under construction, signed chip contracts, and infrastructure investments that rival the interstate highway system.

The wealth this creates is enormous. AMD and NVIDIA are projecting 55-75% profit margins. Companies are seeing 20-40% productivity gains. Stock prices are soaring 200-400%.

And right now, that wealth is flowing almost entirely to shareholders.

While workers lose 45,000-50,000 jobs per month. While communities sacrifice billions in tax revenue for 100 permanent jobs. While companies earning record profits receive taxpayer subsidies to build the infrastructure that will automate away more jobs.

There’s virtually no regulation preventing any of this.

The question isn’t whether AI infrastructure is a bubble. The question is whether workers and communities will organize to demand their share of the wealth it creates—or whether they’ll watch from the sidelines while shareholders get rich.

What Comes Next

This article examined the macro reality: AI infrastructure is real, massive, and concentrating wealth at the top.

But what do YOU do about it? That depends on who you are:

For Workers:

Coming in Part 2: “What Workers Need to Know: Your 18-Month AI Survival Plan”

A practical guide covering:

Honest automation risk assessment for your role
Four pivot strategies with real examples
The two-track approach: protect yourself individually while pushing for systemic change
18-24 month transition timeline with budget
Resources to get started this month

Because by the time you’re laid off, it’s too late to start preparing.

For Communities:

Coming in Part 3: “How to Not Get Fleeced by Data Centers: A Community Guide”

What local governments need to know:

Why Meta’s El Paso deal is a terrible model
The hidden costs communities pay
What to demand in negotiations
How to protect ratepayers from subsidizing corporate infrastructure
Case studies of communities that said no—and won

Because $550 million in tax breaks for 100 jobs is not economic development.

The week this article was finalized, we saw: Saudi AI deals worth billions, Microsoft-Anthropic compute partnership, Target confirming enterprise AI use, and Google’s Gemini 3 crushing benchmarks. Each development validates everything documented above. The infrastructure spending is accelerating. The enterprise adoption is expanding. The competition is intensifying. And American workers are still losing 50,000 jobs per month with virtually no regulation or support.

The Choice

You’ve seen the numbers. You understand who’s winning and who’s losing.

Now what?

You can ignore this and hope it doesn’t affect you. Or you can prepare strategically while pushing for systemic change.

The infrastructure is being built right now. The jobs are being eliminated right now. The wealth is being concentrated right now.

The only question is whether you’ll be positioned to navigate it—and whether you’ll fight to make it fair.

This is Part 1 of a 3-part series on AI infrastructure and its impact on workers and communities. Parts 2 and 3 will be published in the coming weeks on The Open Record.

Subscribe to receive Parts 2 and 3, plus our weekly “Under the Radar” career intelligence newsletter identifying emerging opportunities before they saturate.